Introduction
The Microarray Data Analysis App (MADApp) is a shiny application for
microarrays data analysis. It offers various modules, for image
acquisition, editing, label loading, background correction with multiple
available methods, false positive detection of the spots,
post-processing, array visualization and time series plotting. It is
designed using two-column layout consisting of a structured side panel
and a main panel with a visualisation and plotting area. The analysis
workflow follows conventional microarray processing steps, with each
step organized within one of the seven tabs.
Tab 1: Image Processing Module
The module facilitates the uploading of images and adjustment prior
to analysis (Figure 1).
Date of test and ID
Date of Test allows you to set a custom timestamp
for your measurements. If no date is specified, the system automatically
usese current date. When working with multiple microarrays that share
the same timestamp, you can assign a unique ID to
distinguish between them. Both date and ID must be unique to
successfully add the measurements to intensity data.
Load Image or Sample
To begin analysis, load an image file or use the provided sample
image to explore the application’s features. Once loaded, the image
appears in the main plot where you can zoom, rotate, and adjust the grid
before proceeding with analysis. The application supports standard image
formats including PNG, JPEG and TIFF files.
Rotate Image
The rotation panel enables fine rotation adjustment up to 45 degrees
in both directions (clockwise and counterclockwise). For precise
control, select the slider and use your keyboard arrow keys to make
small adjustments. The image can be also rotated 90 degrees, flipped
horizontally or vertically, using respective buttons:
| 90 degrees counterclockwise |
-90 |
| 90 degrees clockwise |
+90 |
| Flip horizontally |
FH |
| Flip vertically |
FV |
Grid parameters
In the Grid parameters panel, Number of
rows and Number of columns in the grid can be
adjusted. Additionally, there are two grid modes,
Rectangular and Parallelogram,
covering different types of arrays.
After loading the image, click and drag on the main plot to define
the grid area. Double-click within the region of interest (ROI) to zoom
to that region. Use Reset button to return to the
original zoom level and clear the selection. You can adjust grid
parameters before or during the selection process. The following
screenshot shows an example of properly pre-processed and gridded
image.
By clicking Apply Grid button the selected region is
cropped and the analysis proceeds to the next step. To undo all changes
and start the from scratch analysis, press Reset
button.
Tab 2: Annotation
The annotation interface displays the cropped ROI overlaid with the
previously configured grid within the main panel. You can import
existing spot assignment data stored in CSV format files, modify
pre-existing spot assignments, or generate new array annotations from
scratch.
Load Label File
A comma separated value file (CSV file) containing the annotation
information may be loaded here. A annotation data is optional for basic
analysis, but certain advanced analysis functions in post-processing
module and time series analysis require annotation metadata for full
functionality. The annotation file must be CSV file with two columns
with no header row. The first column contains the cell name (e.g. A11),
and the second contains the analyte name. The application default grid
size is 30 x 30, so cell names can be any letter / two letters from A to
AD (single: A, B, C … or double letters: AA, AB …) combined with numbers
1 through 30. The grid size can be further increased in settings
menu.
Remove Labels
By clicking on this button annotation metadata can be removed from
the analysis.
Save Labels
Label will be saved in a CSV File and the file can be later used in
another analysis.
Add Analyte
If the users wants to create a analyte list from scratch or want to
extend the analyte list with new custom name, this can be done by
assigning analyte name and adding it to the list by clicking on
Add Analyte button. To define an analyte on the grid,
first the analyte has to be selected in the Select Analyte element and
click the desired field in the main plot.
When the analyte name is selected in the Select
Analyte box, the analyte marker(s) will be shown on the grid
with green color in the cells where this analyte is present.
Hide Analyte Markers
Analyte markers on the grid can also be hidden, to have better
overview of the spots in the original image. This can be done by
clicking Hide analyte markers. To show them again,
Show analyte markers can be selected.
When the annotation is finished, or if one chooses to skip this step,
one has to click Apply Segmentation and the analysis
will proceed to the next step.
Tab 3: Background Correction
In the background correction tab, one of the threshold methods is
carried out to improve signal to noise ratio and extract the signal from
the background.
Color image?
The application is designed to perform background correction and
extract intensities from grayscale images. Therefore, when working with
a standard three-channel colour image, either a specific channel must be
selected, or the image must first be converted to grayscale. For this,
two approaches can be employed, luminance and grayscale. In the
luminance method, the red, green and blue (RGB) channels are weighted
and summed according to the formula 0.2126 x R + 0.7152 x G + 0.0722 x
B. In the grayscale method, the channels are equally weighted and summed
using the formula 1/3 x R + 1/3 x G + 1/3 x B.
Spot Type
Depending on the type of microarray used, one of the two type of spot
detection mode can be selected: dark spot on a light background or light
spot on a dark background.
Select threshold method and apply
The background that is subtracted from the image can be computed
using two algorithms: Otsu and quantile. Otsu is an automatic image
thresholding method that selects an optimal threshold based on the
discriminant criterion. This threshold can be computed globally, using
information from all cells, or locally, for each cell individually. The
second method, quantile, is a semi-automatic method that computes an
empirical quantile from the intensity values of each cell. The computed
threshold is used to separate the signal from the background in each
cell.
By clicking Apply Background Correction the selected
threshold method is applied to each cell of the microarray grid. After
the thresholding method is applied, two plots and a table containing
information are displayed in the main panel of the third tab (Figure 2).
The plot on the left side shows the pixels above the threshold, while
the plot on the right displays the signal after background subtraction.
Additionally, the displayed table shows the calculated intensities
alongside the threshold data for each cell of the array.
To proceed to the post processing step Proceed to
Post-Processing has to be selected.
Tab 4: Post-processing
Optional: False Positive Spots Detection
In the post-processing tab, the cells with false positive spot
(produced by either artifacts or spot leakage in the adjacent cell) will
be automatically detected by the algorithm and marked on the screen with
purple dot. The algorithm calculates the ratio of pixels above the
background to the total number of pixels within each cell and compares
this ratio to a user-defined threshold.The default threshold value is
0.2, meaning that any cell with less than 20 percent valid pixels will
be marked as false positive. The process is interactive, allowing the
user to adjust the threshold to determine the most appropriate value.
Additionally, each cell can be manually adjusted after visual
inspection. For these field to be excluded from the analysis,
Exclude false positives has to be selected.
Optinal: Scaling and signal refinement
Another optional step is scaling and signal refinement, in which the
extracted intensities are transformed to fit within a range between zero
(indicating no signal), and one (indication the highest intensity)
(Figure 4). The 0.9 empirical quantile of background (empty) cells,
specified in Choose Empty IDs representing IDs of
Negative control or Empty Fields, is used to filter out noise and
artefacts, while the 0.1 empirical quantile of positive control cells,
specified in Choose Positive Control IDs representing
IDs of Positive control Fields. This is achieve using a modified min-max
scaling algorithm, x’ = (x - MIN) / MAX, where MIN is a 0.9 empirical
qunatile of one or multiple empty cells, and MAX is a 0.1 empirical
quantile of on or more cells with the highest intensity. Transformed
intensities below zero are excluded, while those above one are set to
one. In order to use and apply this changes Scale data
has to be checked.
Finally, the post-processed data can be added to the intensity data
table, by clicking Add to Data and the user can proceed
with the analysis of another array. Additionally, a HTML report can also
be generated (Generate Report) that summarises the key
analysis results and includes details of all pre-processing steps, the
parameters applied and system information, ensuring full transparency
and reproducibility.
Clicking Show Intensity Data changes the active tab
to Intensity Data.
Tab 5: Intensity Data
The intensity data tab displays a table containing intensity data
extracted from the microarrays. This step is also an optional starting
step, as previously exported intensity data can be loaded for use in the
next two steps. Additionally, the intensity data can be saved as a CSV
file or deleted, in which case the analysis can be started from the
beginning using a fresh application state. The table contains the
following information: date and ID of the analysis, file name, colour
conversion mode, thresholding method, grid cell, analyte name, mean and
median measurements, and threshold used for background correction. Each
table column can be sorted and filtered.
Load Intensity Data
Load a CSV file containing Intensity Data from previous analysis.
Delete Intensity Data
This button can be used to delete the current intensity data and
start over.
Save Intensity Data
The generated intensity data can be saved in CSV file.
Tab 6: Array Visualization Overview
In the array visualisation tab (Figure 6), the intensity matrices are
used to visualise a measurements in a checkerboard-like table, showing
the mean or median intensity, or threshold for each cell. The date and
measurement ID of the microarray can be specified in the sidebar,
together with an analyte to mark the cell in which it was used.
Additionally, the heatmap feature can be used to colour code the table
and indicate cells with high intensity.
Tab 7: Time Series
In the final tab, Time Series Visualization (Figure 7), one can
specify a time period and an analyte to produce a visualisation of
analyte intensity over time in the cells where the analyte was used, as
well as the average analyte intensity across all cells.
